1. Field of the Invention
This invention relates in general to an improved use of microwave irradiation for heating fluidized bed reactors. In a more particular aspect, it relates to an improved apparatus and process of microwave heating for producing polycrystalline silicon utilizing a fluidized bed in which silicon source gas is thermally decomposed or reduced to deposit silicon on silicon particles.
2. Description of the Prior Art
High-purity silicon with a polycrystalline structure (i.e., "polycrystalline silicon" or "polysilicon"), which is a base material for semiconductor devices and solar cells, is produced by thermal decomposition or hydrogen reduction of silicon source gas. This process is called chemical vapor deposition (CVD). Throughout the world, semiconductor grade polysilicon has been produced by the Siemens reactor. In that batchwise CVD reactor, slim rods of polysilicon are heated by electric current, and the heated rods are exposed to a gaseous mixture of hydrogen and silicon source gas. Recently, attempts have been made to use fluidized beds as an improved CVD reactor system for the bulk production of the high-purity silicon. Fluidized bed reactors are a useful means for the high exposure of solid surfaces to the reactant gases, which provides economy of operation. If a silicon source gas is passed through a fluidized bed of polysilicon particles, abbreviated as "silicon particles" hereinafter, elementary is silicon deposited on their surfaces, and thus the particles grow in size. In addition, another outstanding feature of the fluidized bed system is that the polysilicon product is obtained in the form of granules that are approximately spherical. Being free flowing, this granular polysilicon is transported and handled readily, while rod-like product from the Siemens reactor has to be broken into chunks for transforming polysilicon into single crystal by the Czochralski method. Therefore, the granular polysilicon is essential to the continuous growth of single crystal by a modified Czochralski puller.
Use of the fluidized bed reactor is not without problems in spite of its advantages. When fluidized beds are used for the production of polysilicon, silicon particles in the bed are vulnerable to the agglomeration or sintering of particles due to prolonged contact between them, effectively glued together by the silicon being deposited on their surfaces at a high temperature of about 1000.degree. C. The agglomerated or sintered particles can be grown into clusters that have tendency to settle to the bottom of the fluidizing bed. Then, difficulties in operation are encountered and effective surface areas for silicon deposition are greatly reduced. This problem is deeply related with particle size and temperature. According to the Geldart's classification of particles (see Kunii and Levenspiel, Fluidization Engineering, 2nd ed., Butterworth-Heinemann, 1991, pp. 75-79), the agglomeration or sintering is prominent in fluidizing fine particles (Group C), and can greatly be reduced with increasing particle size above about 40 microns. If fluidizing particles are larger than about 100 microns (classified as Group B or D particles), interparticle cohesive forces are negligible when compared with viscous forces exerted on the particles by fluidizing gas. Thus, the prolonged contact between silicon particles is expected to be prevented by the increase of particle size as well as by proper reactor design to distribute fluidizing gas uniformly. On the other hand, sintering of such silicon particles should be more related with temperature. In a publication (Ceramic Fabrication Processes, ed. W. D. Kingery, MIT Press, Cambridge, Mass., 1958, pp. 120-131) the temperature dependence of the rate of sintering in ceramics, K, is described to be significant as K.varies. exp[-Q/RT] where Q, R and T are the energy of activation, the gas law constant and temperature, respectively. In addition, the temperature dependence of the rate of silicon deposition is of the same form. Consequently, the contact of silicon particles to and the silicon deposition near reactor walls maintained over the required CVD temperature should lead to more probable bridging between particles, together .with the reinforcement by silicon deposition at the bridges, then to a pendular state of agglomerated particles and finally to the formation of clusters. In case of heating reactor walls by resistance heaters, which has been widely used, the walls are inevitably the hottest area within the reactor. Therefore, the silicon deposition near the hot reactor walls is not recommendable in preventing the problem of particle sintering.
Using resistance heaters for fluidized bed CVD reactor also causes the problems of heavy wall deposition, contamination of polysilicon product, and difficulties in material selection and reactor design. To maintain the deposition temperature within the fluidized bed of silicon particles, the reactor wall should be heated much higher than the deposition temperature if the resistance heaters are used around the wall. Then, the rate of the silicon deposition onto the inner wall of the reactor is faster than that onto silicon granules due to the high temperature dependence of silicon deposition. This process cannot therefore be carried out continuously but has to be terminated periodically for replacing or cleaning the reactor wall. Moreover, when using a quartz reactor that is one of the most recommendable material for producing high-purity silicon, the reactor becomes very susceptible to thermal shock because of the extremely large difference between the thermal expansion of the quartz reactor wall and that of the silicon layer deposited thereon. Then, the heavy wall deposition usually results in that the quartz reactor breaks during cooling before cleaning or may even break earlier. Use of a graphite liner, the interior wall which is coated with silicon carbide by an initial silicon deposition reaction, inside a reactor chamber was also proposed to prevent wall breakage (see U.S. Pat. No. 4,092,446). Because silicon and silicon carbide are similar in thermal expansion, the liner can effectively be used for prolonged CVD operation. However, the silicon product from this kind of reactor would not be free from the contamination of carbon impurity. Furthermore, accumulation of heavy silicon deposition on the liner wall should obviously entail the periodic interruption of the CVD operation. Therefore, an efficient method for heating the fluidized bed other than the wall heating is required to use effectively a fluidized bed reactor for producing polysilicon granules. Heating of the silicon particles may be carried out by means of electrodes installed within the fluidized bed or by continuously circulating the particles through a reactor and through a separate heater. These heating methods involve the use of complicated additional apparatus within the fluidized bed itself and/or involve the continuous circulation of the silicon particles into and out of the reactor, and thus they increase the complexity in design and operation. Furthermore, the increase of contact between silicon particles and the surfaces of the electrode wall or of the separate heater with the required piping for particle circulation should apparently lead to the increase of contamination with impurity in the silicon product.
To minimize the problem related with wall heating, a recirculating fluidized bed reactor was proposed as discussed in U.S. Pat. Nos. 4,416,913 and 4,992,245 and Japanese Pat. KOKAI No. 2-30611 (1990). The reactor is characterized by peripheral heating zone annulus containing downwardly flowing silicon particles, which are heated by outer heating zone wall, i.e., the reactor wall, and are then transferred to an inner reaction zone. The particles enter the heating zone annulus at an upper inlet and exit the annulus at a lower outlet. The CVD reaction mainly proceeds in the inner reaction zone through which a silicon source gas rises upwardly. The mass flow rate of falling particles through the heating zone and the temperature difference between the outer annulus wall and these particles should be high enough to supply the required heat to the annulus. It is then difficult to supply upwardly sufficient amount of a carrier gas such as hydrogen for preventing leakage of the silicon source gas from the reaction zone through both ends of the heating zone, while maintaining the high downward mass flow of the silicon particles to be heated in the heating zone. The flow rate of the carrier gas would be at most that required for the incipient fluidization of the falling particles. Thus, the introduction of the silicon source entrained with the falling silicon particles and its gaseous diffusion from the reaction zone into the annular space cannot avoided by such amount of uprising carrier gas. Although the amount of the silicon source entrained into the heating zone is not significant, a fraction of it should naturally be decomposed to silicon on the outer annulus walls which must be considerably hotter than the particles. Therefore, the proposed circulation reactor can reduce but cannot prevent the problem of the deposition at the outer annulus wall heated outside. Besides, the convective heat swept and lost by the carrier gas uprising the heating zone is considerable. It is also required to employ some type of driving force such as a pulsed gas jet to promote the introduction of the heated particles from the lower outlet of the heating zone to inside the reaction zone. Then, the heat loss due to the carrier gas and the pulsed gas jet should be made up by increasing the wall temperature. The requirements of the high temperature at the reactor wall and the low degree of fluidization within the heating zone can yield the agglomeration of particles near the heated walls. In addition, due to the lower outlet of the heating zone, the gas distribution means for introducing a silicon source should always contact with the heated particles. This naturally leads to the deposition of silicon and the formation of crust on the distribution means unless the means is sufficiently cooled below an incipient decomposition temperature of the silicon source. It is apparent that such cooling reduces the heat transfer from the heating zone to the heating zone. Therefore, a need exists for an improved type of the fluidized bed reactor with separated heating zone.
To overcome the disadvantages of supplying heat to the wall of the fluidized bed reactor, it has been proposed to heat silicon particles in a CVD reactor by the irradiation with an electromagnetic wave such as microwaves that cover the range of frequencies from 50 MHz to 300 GHz. Since silicon is a highly microwave absorptive material, the microwave can be used effectively in the fluidized bed CVD process. U.S. Pat. No. 4,416,913 describes a possible use microwave heating to keep silicon particles in a rising particle reaction column hotter than the chamber walls so that CVD on the walls is reduced or avoided. It is recirculating rising particle reactor where arising particle reaction zone column is installed within an annulus reservoir arranged surrounding the reaction column. In this column the gas velocity of silicon source gases must be sufficient to lift, transport, and eject all silicon particles smaller than a predetermined size while those which have grown to a larger size fall through the rising gas stream and are extracted from the base of the reactor. It is notable that the CVD reactor proposed in the patent is not the fluidized bed type reactor in which net solids flow is zero or nearly zero and most of the solid particles fluidized by rising gas bubbles contact with each other. The rising particle reactor is an example of a dilute-phase (or lean-phase) solid-gas system because the volume fraction of solid is much lower than in the dense-phase fluidized bed reactor. The rising particle reactor thus requires excessively high velocity of silicon source gas to overcome gravitational force of silicon particles and to maintain upward solids flow (see Perry et al, "Chemical Engineers' Handbook", 5th Ed., pp 20-64, McGraw-Hill, Inc. 1973). Under such a dilute-phase system the microwave heating of silicon particles is practically difficult due to light loading factor, i.e., low volume fraction of silicon particles (microwave absorptive load). Since most silicon particles are uprising separately from each other, heat transfer by solid-solid contact and by radiation between them is greatly limited in the rising particle reactor. In addition, cooling of the heated particles by the high velocity of silicon source gas also requires high-power microwave energy, because the gas cannot be preheated sufficiently before feeding due to the initiation of CVD within a preheater. Therefore, the feasibility of the microwave heating is not expected in such a dilute-phase CVD system.
More improved use of the microwave heating for the production of polysilicon granules has been disclosed in U.S. Pat. 4,900,411. Following the description in the patent, microwaves are introduced into a lower reaction zone of the fluidized bed reactor. In the reaction zone silicon particles are fluidized by the reaction gas and are heated by direct irradiation of microwaves. Compared with the above-mentioned particle rising reactor, this dense-phase fluidized bed reactor is more effective in microwave heating because of the higher volume fraction of silicon particles and of the lower gas velocity within the reactor. The heating method was observed to keep the reactor walls not hotter than that of silicon particles because quartz is transparent to microwaves in the range of CVD temperature. In spite of these advantages the method includes some undesirable manipulations for CVD reaction that are requisite to the direct irradiation of microwaves into the reaction zone. When microwaves penetrate into the reaction zone through quartz reactor walls, the silicon particles next to the walls absorb most of the microwave energy because the penetration depth of microwaves decreases with temperature. Thus, the inner walls of the quartz reactor in direct contact with the irradiated silicon particles should be kept as hot as these particles. Therefore silicon deposition on the inner walls by the reaction gas occurs with the same rate as on heated silicon particles. If the silicon layer deposited on the walls through which microwaves penetrate becomes sufficiently thick, the layer itself would absorb most of the microwave energy. This would lead to accumulation of heat inside the irradiated layer and then to an accelerated temperature increase even above the melting temperature of silicon. In this case the advantage of microwave heating vanishes. To prevent this problem the referenced patent describes a gas cooling outside the reactor wall without disturbing the microwave irradiation into the reaction zone. It also describes that such a wall cooling is important in CVD operations at 700.degree. C. if monosilane is used as a silicon source gas. However, such wall cooling naturally increases heat loss and thus requires high microwave power to maintain the reaction temperature. In addition, insulation surrounding the reactor walls is impossible. Thus, the gas cooling greatly increases power consumption, and reduces the feasibility of microwave heating. If CVD operations are executed with trichlorosilane as a silicon source gas at above 900.degree. C., the wall deposition should become more probable in spite of cooling the reactor walls. This is attributed to the characteristics that: trichlorosilane can be decomposed to silicon both on the microwave irradiated silicon particles and on the reactor wall unless the wall is cooled under 400.degree. C. without any selectivity for the temperature of solid surfaces, while the pyrolysis of silane occurs selectively on hotter silicon particles than on cooled reactor walls. Therefore, gas cooling of the reactor wall should become more serious but difficult in operation with the increase of the reaction temperature. Furthermore, to save energy, insulation around the reactor walls is impossible. Besides the wall cooling, the patent requires cooling a gas distributing means supporting the reaction zone by introducing a coolant fluid into the means to prevent substantial deposition of silicon on it. Then, the referenced CVD process is shown to be based on simultaneous heating and cooling of the reaction zone, i.e., direct microwave heating of the reaction zone together with cooling by coolant fluids of the solid surfaces encompassing the reactor, which reduces the effect of microwave heating and increases energy consumption as well as difficulties in operation.